U.S. patent application number 11/818888 was filed with the patent office on 2008-12-18 for network interface selection using historical connection information.
This patent application is currently assigned to Microsoft Corporation. Invention is credited to Deepak Bansal, Sumit Garg, Murari Sridharan, David Thaler.
Application Number | 20080310419 11/818888 |
Document ID | / |
Family ID | 40132250 |
Filed Date | 2008-12-18 |
United States Patent
Application |
20080310419 |
Kind Code |
A1 |
Bansal; Deepak ; et
al. |
December 18, 2008 |
Network interface selection using historical connection
information
Abstract
A computer may be capable of communicating with devices via
network through multiple interfaces. When a new connection is to be
established, the computer may select an interface to use for the
connection. In some embodiments, the interface may be selected
based on historical connection information representing the
performance of previous network connections that the computer has
established through the interfaces.
Inventors: |
Bansal; Deepak; (Redmond,
WA) ; Garg; Sumit; (Seattle, WA) ; Sridharan;
Murari; (Sammamish, WA) ; Thaler; David;
(Redmond, WA) |
Correspondence
Address: |
WOLF GREENFIELD (Microsoft Corporation);C/O WOLF, GREENFIELD & SACKS, P.C.
600 ATLANTIC AVENUE
BOSTON
MA
02210-2206
US
|
Assignee: |
Microsoft Corporation
Redmond
WA
|
Family ID: |
40132250 |
Appl. No.: |
11/818888 |
Filed: |
June 15, 2007 |
Current U.S.
Class: |
370/395.6 |
Current CPC
Class: |
H04L 45/124 20130101;
H04L 45/00 20130101; H04L 45/12 20130101; H04L 45/22 20130101 |
Class at
Publication: |
370/395.6 |
International
Class: |
H04L 12/56 20060101
H04L012/56 |
Claims
1. In a computing environment including a first computer configured
to communicate via a network through a first network interface and
a second network interface, the network enabling the first computer
to communicate with a plurality of other computers via the network,
a method comprising acts of: (A) storing, by the first computer,
first historical network connection information representing a
performance of a first network connection between the first
computer and at least one of the plurality of other computers
through the first network interface; (B) storing, by the first
computer, second historical network connection information
representing a performance of a second network connection between
the first computer and at least one of the plurality of other
computers through the second network interface; (C) selecting,
based on at least the first and second historical network
connection information, a selected network interface from among the
first and second network interfaces to use for a third network
connection; and (D) based on the act (C), establishing the third
network connection between the first computer and a second computer
from among the plurality of other computers such that third network
connection communication passes through the selected network
interface.
2. The method of claim 1, wherein the first network interface and
the second network interface operate according to different network
communication protocols.
3. The method of claim 1, wherein the first network interface and
the second network interface provide access to the network through
different types of communication media.
4. The method of claim 3, wherein the different types of
communication media comprise a wireless medium and a wired
medium.
5. The method of claim 1, wherein the act (C) is performed by the
first computer and the first computer initiates the act (D) in
response to the act (C).
6. The method of claim 1, wherein the first network connection is
between the first computer and a third computer from among the
plurality of other computers, wherein a network address of the
third computer is similar to a network address of the second
computer, and wherein the selected network interface is selected to
be the first network interface based in part on a similarity of the
network addresses of the second and third computers.
7. The method of claim 6, wherein the address of the third computer
and the address of the second computer have a same prefix.
8. The method of claim 6, wherein the third computer has a physical
location that is close to a physical location of the second
computer.
9. The method of claim 1, wherein at least a portion of the first
historical network connection information represents a hop count
for the first network connection, and wherein the selected network
interface is selected to be the first network interface based at
least in part on the hop count.
10. The method of claim 1, wherein at least a portion of the first
historical network connection information represents a round-trip
time for the first network connection, and wherein the selected
network interface is selected to be the first network interface
based at least in part on the round-trip time.
11. The method of claim 1, wherein the first historical network
connection information represents a data rate for the first network
connection, and wherein the selected network interface is selected
to be the first network interface based at least in part on the
data rate.
12. The method of claim 1, wherein the first historical network
connection information represents a packet loss rate for the first
network connection, and wherein the selected network interface is
selected to be the first network interface based at least in part
on the packet loss rate.
13. The method of claim 1, wherein the selected network interface
provides a higher overall connection data rate than a non-selected
network interface of the first and second network interfaces would
provide.
14. The method of claim 1, wherein the act (A) is performed prior
to terminating the first network connection and the act (C) is
performed after terminating the first network connection.
15. In a computing environment including a first computer
configured to communicate via a network through a first network
interface and a second network interface, the network enabling the
first computer to communicate with a plurality of other computers
via the network, a computer readable medium having instructions
which, when executed, perform a method comprising acts of: (A)
storing, by the first computer, historical network connection
information representing a performance of a plurality of network
connection between the first computer and at least one of the
plurality of other computers through the plurality of network
interfaces; (B) selecting, based on at least the historical network
connection information, a selected network interface from among the
plurality of network interfaces to use for a network connection
between the first computer and at least one of the plurality of
other computers; and (C) based on the act (B), establishing the
network connection between the first computer and a second computer
from among the plurality of other computers such that communication
between the first computer and the second computer passes through
the selected network interface.
16. The computer readable medium of claim 15, wherein (A) is
performed such that the historical network connection information
is stored while at least one of the plurality of network
connections is active.
17. A system, comprising: a computer; a first network interface
that enables the computer to communicate with a plurality of other
computers via a network; and a second network interface that
enables the computer to communicate with a plurality of other
computers via the network; wherein the computer comprises: at least
one storage device storing first historical network connection
information representing a performance parameter of a first network
connection between the computer and at least one of the plurality
of other computers through the first network interface, and second
historical network connection information representing a
performance parameter of a second network connection between the
computer and at least one of the plurality of other computers
through the second network interface; a processor that determines,
based on the first and second historical information, which of the
first and second network interfaces to use for a third network
connection to be established with at least one of the plurality of
other computers.
18. The computer of claim 17, wherein the first and second network
interfaces are part of the computer.
19. The computer of claim 17, wherein the computer further
comprises: at least one wired bus coupled to the processor; wherein
the first network interface is a network adapter in communication
with the processor via the at least one wired bus.
20. The computer of claim 17, wherein the first network interface
is a wired network adapter and the second network interface is a
wireless network adapter.
Description
BACKGROUND
[0001] 1. Technical Field
[0002] The techniques described herein relate to network
communication using a computer.
[0003] 2. Discussion of Related Art
[0004] Computers can be configured to communicate with one another
over a network connection. The details of a network connection may
not be readily apparent to the user. For example, a computer may
communicate with a server that could be located nearby or in a
different state or country. Communication may take place over
different types of media, such as telephone lines, cable lines,
fiber optics or radio waves, for example. If the network
communication takes place over the Internet, communication may be
facilitated by an Internet Service Provider (ISP), which may route
the communications through any one of a variety of different
network paths.
[0005] Some computers may have a variety of different options for
how they will access a computer network. For example, a computer
may access a network through an Ethernet connection which provides
a wired network interface or a wireless network interface card
(NIC) that provides a wireless network interface. When a computer
has more than one network interface it may choose which interface
to use for a connection. One prior technique for choosing a network
interface is to select an available interface that has the highest
interface data rate. The computer's interfaces may advertise the
data rate that they are capable of providing and the computer may
select the interface based on the advertised data rates.
SUMMARY
[0006] In one aspect, when a computer seeks to establish a new
network connection, a network interface may be chosen for the
network connection based on the performance of previous network
connections through the available interfaces. Applicants have
appreciated that, although one interface may have a higher data
rate than another interface, in some embodiments using the slower
interface may provide better overall performance for the network
connection. Historical information about previous connections
through multiple interfaces may be used to predict which interface
will provide the best overall performance for a new connection. In
some embodiments, such historical information may be stored by the
computer itself. As the computer makes connections to other
computers, one or more performance-related parameters of those
connections may be stored and associated with the interface through
which they were established. By way of example, such information
may include the round-trip time, the connection data rate, the
connection loss rate, the hop count for the connection and/or any
other suitable information. When a new connection is desired to be
established, the computer may access some or all of this stored
information to determine which interface to use. As a result,
improved connection performance may be achieved by selecting an
interface that is most likely to provide the best connection
performance.
[0007] One embodiment relates to a method for use in a computing
environment that includes a first computer configured to
communicate via a network through a first network interface and a
second network interface. The network may enable the first computer
to communicate with a plurality of other computers via the network.
The first computer may store first and second historical network
connection information. The first historical network connection
information may represent a performance of a first network
connection between the first computer and at least one of the
plurality of other computers through the first network interface.
The second historical network connection information may represent
a performance of a second network connection between the first
computer and at least one of the plurality of other computers
through the second network interface. Based on at least the first
and second historical network connection information, a network
interface may be selected to use for a third network connection
from among the first and second network interfaces. Based on the
interface selected, a third network connection may be established
between the first computer and a second computer from among the
plurality of other computers such that third network connection
communication passes through the selected network interface.
[0008] Another embodiment relates to a computer readable medium for
use in a computing environment that includes a first computer
configured to communicate via a network through a first network
interface and a second network interface. The network may enable
the first computer to communicate with a plurality of other
computers via the network. The computer readable medium may have
instructions recorded thereon which, when executed, perform a
method that includes several acts. For example, in response to the
execution of the instructions, the computer may store historical
network connection information representing a performance of a
plurality of network connection between the first computer and at
least one of the plurality of other computers through the plurality
of network interfaces. A network interface may be selected from
among the plurality of network interfaces to use for a network
connection between the first computer and at least one of the
plurality of other computers. Based on the interface selected, a
network connection may be established between the first computer
and a second computer from among the plurality of other computers
such that communication between the first computer and the second
computer passes through the selected network interface.
[0009] Yet another embodiment relates to a system that includes a
computer and also includes first and second network interfaces that
enable the computer to communicate with a plurality of other
computers via a network. The computer may include at least one
storage device storing first historical network connection
information representing a performance parameter of a first network
connection between the computer and at least one of the plurality
of other computers through the first network interface, and second
historical network connection information representing a
performance parameter of a second network connection between the
computer and at least one of the plurality of other computers
through the second network interface. The computer may also include
a processor that determines, based on the first and second
historical information, which of the first and second network
interfaces to use for a third network connection to be established
with at least one of the plurality of other computers.
[0010] This Summary is provided to introduce a selection of
concepts in a simplified form that are further described below in
the Detailed Description. This Summary is not intended to identify
key features or essential features of the claimed subject matter,
nor is it intended to be used to limit the scope of the claimed
subject matter.
BRIEF DESCRIPTION OF DRAWINGS
[0011] In the drawings, each identical or nearly identical
component that is illustrated in various figures is represented by
a like numeral. For purposes of clarity, not every component may be
labeled in every drawing. In the drawings:
[0012] FIG. 1 illustrates a computing environment in which
embodiments may be implemented;
[0013] FIG. 2 is a flowchart illustrating a method of communicating
over a network, according to some embodiments; and
[0014] FIG. 3 illustrates an exemplary computer on which
embodiments may be implemented.
DETAILED DESCRIPTION
[0015] Applicants have appreciated that prior techniques for
selecting a network interface suffer from disadvantages. As
discussed above, one prior technique for selecting a network
interface is to choose an available interface that advertises that
it can provide the highest data rate. For example, an Ethernet
network adapter interface may advertise that it is available and
capable of providing a high data rate, and a cellular data network
adapter interface may advertise that it is available and capable of
providing a lower data rate. According to prior techniques for
selecting the best available interface, the Ethernet interface
would be selected because it advertises that it is capable of
providing better performance than the cellular data interface.
However, in some circumstances, connecting via the cellular data
interface may actually provide better performance, as the interface
speed itself does not take into account factors beyond the
interface that may affect network performance. For example, the
Ethernet connection may provide a connection to a server of an
internet service provider (ISP). However, if the ISP is currently
unable to provide a high quality of service, a connection made
through the Ethernet connection may be relatively slow. The
computer may experience better performance by connecting using the
cellular data interface which may provide a network connection
through a different ISP that may provide better service.
Accordingly, Applicants have appreciated that factors besides the
interface speed may limit the performance that can be achieved by a
connection. Thus, in some embodiments, different and/or additional
criteria may be used for selecting the interface to be used for a
connection. For example, the performance of previous connections as
a whole may be taken into account rather than just the capabilities
of a network interface.
[0016] In some embodiments, an interface may be selected based on
historical information about the performance of previous
connections through the available interfaces. Such historical
information may include any of a variety of factors, such as the
round-trip time, the connection data rate, the connection loss
rate, the hop count for the connection and/or any other suitable
information. The computer may record historical information about
such connection parameters when connections are established, and
may access this historical information to make a decision about
which interface to use for a new connection. Such historical
information may not be limited to the performance of the interfaces
themselves, but may in addition to or instead reflect the
performance of the connection as a whole. If an interface has
recently provided good performance for a network connection to a
particular network location, it may be assumed that the interface
may continue to provide good performance to that network location
and to locations nearby. Under those circumstances, such an
interface may be selected over a faster interface which may have
provided inferior service to a similar network location. Suitable
historical information about previous network connections may be
used to select an interface that is more likely to provide good
performance in the future. In some embodiments, such a solution may
advantageously be implemented by the computer itself based on
historical information about previous network connections.
[0017] FIG. 1 illustrates an example of a computing environment 10
in which embodiments may be implemented. Computing environment 10
includes a first computer 1 and a second computer 2 to which
computer 1 may establish a connection via network 7. Computers 1
and 2 can be computers of any suitable type, and network 7 can be a
network of any suitable type, as the techniques described herein
are not limited in these respects. First computer 1 may be in
communication with a first interface 3 and a second interface 4
which may enable computer 1 to exchange information with devices
via network 7. If computer 1 is connected to computer 2 through
interface 3, communication may take place via a first route 5
through network 7. If computer 1 is connected to computer 2 through
interface 4, communication may take place via a second route 6
through network 7. Routes 5 and 6 may experience different
performance.
[0018] Although only two interfaces are illustrated in FIG. 1, the
techniques described herein are not limited to this example as
additional interfaces may provide access to network 7, and these
techniques may be extended to these additional network interfaces.
A network interface may be part of a computer or separate from a
computer. For example, a computer may have a built-in network
interface in the form of a network interface card (e.g., a built-in
wireless NIC or Ethernet NIC) which is a part of the computer.
Alternatively, an interface may be separate from a computer, and
may communicate with the computer via a local data bus (e.g., a
serial or parallel bus) to provide access to another device via a
network. Any suitable type and configuration of network interface
may be used, as the techniques described herein are not limited in
this respect.
[0019] Computer 1 may subsequently attempt to establish a new
connection to computer 2 through interface 3, interface 4 or some
other interface. In some embodiments, computer 1 may select the
interface to use based on historical information about the
performance of previous connections that have been established
(e.g., via interfaces 3 and 4 in the example of FIG. 1). Such
historical information may enable the computer to take into account
the effect of interface selection upon the route (e.g., route 5 or
6) that is taken through the network. As one example, interface 3
may be capable of providing a much higher data rate than interface
4. Selecting interface 3 might lead to communications being routed
through route 5 through network 7. However, if route 5 is slower
than route 6, then choosing interface 4 may provide better overall
performance than choosing interface 3. In some embodiments,
historical information about previous connections made through
interfaces 3 and 4 may allow the computer to make an educated
prediction about which interface is likely to provide a connection
with better performance.
[0020] Any suitable historical information may be considered. As
one example, the historical information may include information
about the round-trip time (RTT) for a previous network connection.
When computer 1 sends a packet to computer 2, there may be a delay
before computer 2 receives the packet. Once computer 2 receives the
packet, computer 2 may send a response to computer 1. For example,
computer 1 may send an acknowledgment packet to acknowledge the
receipt of the packet sent by computer 1. The RTT may be an amount
of time it takes from the time a communication is sent by computer
1 until a response is received by computer 1. A smaller RTT may be
preferred because a smaller RTT may enable a connection to provide
better performance. The RTT may be affected by a variety of
factors, such as the distance between the computers, the speed of
computer 2 and/or the speed of the route that is taken through the
network.
[0021] As another example, the historical information may include
information about the data rate that has been achieved by a past
connection. As with RTT, the data rate may be affected by a variety
of factors such as the speed of the route that is taken through the
network. As yet another example, the historical information may
include information about the loss rate for a connection. When
sending data over a network, a packet may not always reach its
destination, and sometimes may reach its destination in corrupted
form. When this occurs, the packet is said to be lost, and the
packet may need to be re-transmitted. The loss rate for a
connection represents the rate at which packets are lost for a
given connection. As a further example, the historical information
may include information about the "hop count" which is the number
of steps that are taken through the network for data to reach its
destination. When data is routed through a connection, the data may
need to pass through multiple steps, each of which may include a
device (e.g., switch, router or gateway) which may forward the data
along. If the hop count is relatively large, the data may need to
pass through a large number of routers which may slow down a
connection, and may present a larger chance for the connection to
fail.
[0022] The above are merely examples of parameters that may be
measure to determine performance of a connection, and it should be
appreciated that the invention is not limited to using these, as
any suitable historical connection information may be used for
interface selection.
[0023] FIG. 2 is a flowchart illustrating a method of communicating
over a network, according to some embodiments. As discussed above,
historical connection information may be saved by a computer for
future use in selecting an interface. In act A1, a computer may
store historical connection information about a connection that has
been established through a first interface. This can be done in any
suitable way, for example, when the computer establishes a
connection through the first interface to a network, the computer
may store information about the connection. This information may be
obtained from any source. The computer itself could measure a
connection information and/or could receive connection information
from another device. For example, the connection information could
be received from another device that measures connection
information. The computer could store the historical connection
information locally in memory, or this information could be stored
elsewhere. For example, the information may be stored by the device
that measures the information and obtained by the computer later
prior to selecting an interface. The historical connection
information may include any suitable performance-related
information about the connection, examples of which are discussed
above. In some embodiments, the historical connection information
may be associated with the network address of the device to which
the computer connected, and the network address and historical
connection information may be stored together. In one example, act
A1 may be performed by computer 1 for a connection established to
computer 2 via interface 3. However, this is just an example, as
act A1 may be performed in any suitable way.
[0024] In act A2, a computer may store historical connection
information about a connection that has been established through a
second interface different from the first interface. The historical
connection information may represent performance-related
information about a connection made through the second interface.
In some embodiments, the historical connection information that is
stored in act A2 may be the same type of information stored in the
act A1, thus allowing different connections to be compared based on
the same criteria. Alternatively, different types of historical
connection information may be stored for different connections, as
the techniques described herein are not limited in this respect. As
discussed above with respect to act A1, the historical connection
information may be measured by the computer or by a different
device, and may be stored locally on the computer or in another
location. In one example, act A2 may be performed by computer 1 for
a connection established to computer 2 via interface 4. However,
this is just an example, as act A2 may be performed in any suitable
way.
[0025] As a result, historical connection information may be saved
for connections through different interfaces. In some embodiments,
acts A1 and A2 may be repeated multiple times so that a computer
may build a dataset of historical connection information for
connections to multiple different devices. For example, connection
information may be measured and saved periodically while a
connection is established. The interval between measurements may be
chosen to be approximately equal to the RTT the connection, for
example, and measurements may be also be made when the RTT changes.
However, connection information need not be measured or saved
periodically, and could be performed in a different manner. Any
suitable raw connection measurement and/or derived connection
information may be stored for one or more connection parameters. If
a computer has access to more than two network interfaces,
historical connection information may also be stored for any of
these additional interfaces, if desired.
[0026] At some point, a computer may wish to establish a new
network connection. To select a network interface, the computer may
make a decision in act A3 based at least in part on the historical
connection information saved in acts A1 and A2. Such historical
connection information may be used in a variety of ways. As one
example, the decision may be made based on one type of historical
connection information, such as the best data rate. As another
example, the decision may be made based on a combination of
different historical connection parameters. If multiple parameters
are used, each parameter may be weighted according to its relative
importance to connection quality. An interface may be chosen that
is likely to provide better performance, as measured by the
weighted parameters. The parameters used and the weights give can
be chosen in any suitable way, as the invention is not limited to
any particular parameters or weighting.
[0027] In some embodiments, multiple measurements of a single
connection parameter may be used for interface selection. For
example, a statistical function or other mathematical function may
be performed upon a plurality of measurements of a single parameter
such that one or more mathematically-generated connection
parameters are determined. As one particular example, an average
connection parameter may be determined based upon multiple
measurements of the connection parameter that have been made over
time, such that the most recent measurements for over a particular
time period are used. More recent measurements may be given more
weight than older measurements, for example. However, the
techniques described herein are not limited as to the method of
weighting multiple measurements or the particular function that is
employed to determine such a mathematically-generated
parameter.
[0028] In some embodiments, a subset of the available historical
connection information may be used for the selection of an
interface. For example, if a new connection is to be established to
a computer having a particular network address, historical
connection information may be considered for previous connections
that were made to computers that had similar network addresses, as
the inventors have appreciated that computers with similar
addresses tend to be located in a similar physical location.
Similarity of addresses can be determined in any suitable way, as
the techniques described herein are not limited to any particular
technique for determining similarity. One analogy that can be made
is with respect to telephone numbers, which may have the same area
code for telephone numbers who's "home address" is in the same
area. The same may be true for internet protocol (IP) addresses, as
computers that are located near each other may have addresses with
the same prefix. Thus, there may be a statistical correlation
between IP address prefixes and physical proximity. In some
circumstances, connection parameters may be similar when connecting
to devices that are physically close to one another. As a result,
connections made to devices having similar IP prefixes may be of a
similar quality, and may have similar connection performance
characteristics. Thus, in some embodiments, when a new connection
is desired to be established to a particular IP address, an
interface may be selected based on historical data for connections
to similar IP addresses. In some embodiments, similar IP addresses
may include IP addresses that are exactly the same, which
represents a connection to the same network location. In internet
protocol version four (IPv4), similar addresses may include IP
addresses that have the same first three bytes or the same first
two bytes, for example. In IPv6, similar addresses may include IP
addresses having the same first eight bytes, the same first seven
bytes, the same first six bytes, or the same prefix having any
other suitable length. Any suitable prefix length may be chosen for
computers that have similar IP addresses, as the techniques
described herein are not limited to the particular size of IP
address prefix that is compared.
[0029] In response to selecting an interface in act A3, a
connection may be established in the act A4 such that communication
via the connection takes place through the selected interface. Any
suitable type of connection may be established, as the techniques
described herein are not limited as to the particular protocol that
is used for connections. One example of a suitable connection is a
transmission control protocol (TCP) connection, however other types
of connections may be established.
[0030] For the computing environment illustrated in FIG. 1,
computers 1 and/or 2 may be any suitable types of computers, such
as the computer described in further detail below. For example,
computers 1 and/or 2 may be desktop computers, laptop computers,
cellular telephones, personal digital assistants or any other
suitable type of computing device. Interfaces 3 and 4 may be any
suitable interfaces that enable computer 1 to access network 7.
Interfaces 3 and 4 may be network adapters that are in
communication with a processor of computer 1 via a wired bus, for
example. Examples of suitable interfaces include wired NICs, such
as Ethernet cards, telephone modems, DSL cards, etc. Examples of
suitable interfaces also include wireless NICs, such as NICs
capable of communicating via Bluetooth, IEEE Standard 802.11,
cellular data protocols, and/or any other suitable type of NICs.
Network 7 may be any suitable computer network such as the
Internet, an intranet and/or any other suitable type of network, as
the techniques described herein are not limited in this
respect.
[0031] A computing system will now be described, on which
embodiments of the invention may be implemented. With reference to
FIG. 3, an exemplary system on which embodiments of the invention
may be implemented includes a computing device, such as computing
device 40, which may be a device suitable to function as one or
more of computers 1 and 2, for example. Computing device 40 may
include at least one processing unit 41 and memory 42. Depending on
the exact configuration and type of computing device, memory 42 may
be volatile, non-volatile or some combination of the two. One
possible configuration is illustrated in FIG. 3 by dashed line 43.
Additionally, device 40 may also have additional
features/functionality. Memory 42 is a form of computer-readable
media that may store information such as computer-readable
instructions, which, when executed, implement any of the techniques
described herein.
[0032] Device 40 may include at least some form of computer
readable media. Computer readable media can be any available media
that can be accessed by device 40. By way of example, and not
limitation, computer readable media may comprise computer storage
media and communication media. For example, device 40 may also
include additional storage (removable and/or non-removable)
including, but not limited to, magnetic or optical disks or tape.
Such additional storage is illustrated in FIG. 3 by removable
storage 44 and non-removable storage 45. Computer storage media
includes volatile and nonvolatile, removable and non-removable
media implemented in any method or technology for storage of
information such as computer readable instructions, data
structures, program modules or other data. Memory 42, removable
storage 44 and non-removable storage 45 are all examples of
computer storage media.
[0033] Device 40 may also include output device(s) 46, input
device(s) 47 and/or one or more communication connection(s) 48.
Communication connection(s) 48 may include, for example, a bus
interface that enables device 40 to communicate with a network
interface via any suitable data bus, such as a parallel bus or a
serial bus.
[0034] It should be appreciated that the invention is not limited
to executing on any particular system or group of systems. For
example, embodiments of the invention may run on one device or on a
combination of devices. Also, it should be appreciated that the
invention is not limited to any system or network architecture.
EXAMPLE 1
[0035] One detailed example of an application of the techniques
described herein will now be described. However, this is just an
example, and the invention is not limited to this particular
application, as the techniques described herein may be used in
numerous other applications.
[0036] In some embodiments, an interface may be chosen to achieve
the highest bandwidth at layer four of the OSI network
communication module. Since bandwidth achieved on a particular
route may be measured, past performance may be used to make more
informed decisions about future interface selection. Below is
describe an example of a solution based on end-to-end bandwidth
measurements from the past to make route selection. [0037] 1. Each
off-link route may maintain two bandwidth values: near
(corresponding to bandwidth measurement from TCP connections to
"nearby" hosts) and far bandwidth (corresponding to bandwidth
measurement from TCP connections to hosts which are not "nearby").
For routes to destinations on the same subnet, there may be only
one value of measured bandwidth. [0038] 2. Initially, the near and
far bandwidths may be set to values derived based on interface
speed reported by the NICs. Until sufficient information has been
reported on bandwidth by TCP connections, bandwidth values based on
interface speed are used. Sufficient information may be obtained by
storing connection information for at least 5 connections to 5
different destinations, in some embodiments. [0039] 3. For each TCP
connection, bandwidth obtained may be measured and reported to IP
layer as follows: [0040] a. The bandwidth sample may be taken and
reported to the IP layer by the TCP layer if [0041] i. a) The
sender is efficiently filling the pipe (sends at least 50% of what
he is allowed to send) or b) The sender is congestion limited.
(Loss, delay spikes, reordering etc). [0042] ii. For receiver
bandwidth, samples may be provided if autotuning is enabled or 75%
of receive window we get every RTT--efficient receiver. [0043] iii.
The sample may be passed to the IP layer and tracked on a per-route
basis, separating the measurement into subsets or "buckets," such
as near and far buckets:
[0043] AvgBW=AvgBW+(Sample-AvgBW)/N, [0044] N for each case may be
chosen based on testing and experimentation (e.g., 8). [0045] Note
that this algorithm may not differentiate the direction of
transfer; sender and receiver samples are treated alike and are not
summed. So if the data transfer is happening simultaneously, its
estimates will be more conservative. [0046] 4. For each route,
bandwidth information obtained may be categorized into the near or
the far bucket. To do so, the destination IP address for which the
bandwidth is reported is used. This may done as follows: [0047] a.
All destinations whose IP address match the IP address of the host
under the subnet mask of the host left shifted by 8 (for IPv4) or
16 (for IPv6) may be considered near. [0048] b. For destinations
not meeting criteria (a), RTT value or hop count may be used as
follows. If the RTT to the destination is below a threshold (e.g.,
10 milliseconds) or hop count is less than a threshold (e.g., 6),
the destination is treated as near. Further, for future
connections, all destinations that match a near destination under
the subnet mask of the host left shifted by 8 (for IPv4) or 16 (for
IPv6) may be considered near. To do this, the computer may maintain
a list of near IP address ranges. [0049] c. Any IP address that
does not belong to near or is on subnet is considered far. [0050]
5. For a destination to which a new TCP connection is to be
established, the destination IP may first be categorized into near
vs. far and then various routes from the route lookup that match
may be compared based on average near or far bandwidths for those
routes. The route with the higher average bandwidth may be
selected.
TABLE-US-00001 [0050] Near BW Far BW Route 1 X1 X2 Route 2 Y1
Y2
[0051] For a near destination, it route 1 and 2 are matching
routes, X1 and Y1 may be compared while for a far destination, X2
and Y2 may be compared [0052] 6. The average far bandwidth for a
particular route may be upper bounded by the average near bandwidth
for that route. [0053] 7. To increase stability, hysteresis may be
introduced by bucketizing average bandwidth into few buckets. An
average bandwidth may be categorized into a different bucket only
if it crossed the bucket threshold by over 25%. [0054] 8. After a
period of time (say 2 hrs) if no new bandwidth samples have been
reported for a route, the average bandwidth for that route may be
reset to based on interface speed as in 1.
[0055] The state maintained in section 3 above may purged when a
machine reboots. On sleep/wake-up or media disconnect/re-connect,
the state may be purged if any IP address assigned to the interface
is different (since a different IP address indicates connectivity
to a different network and hence, likely to result in different
performance).
[0056] The above algorithm may load balance TCP connections across
interfaces that fall into the same bucket based on interface speed;
e.g., if a large number of TCP connections start going over one
interface such that the average throughput for each connection is
reduced to below what is achievable over the other interface, then
the above algorithm may automatically detect that and start routing
new TCP connections over the other interface.
[0057] Below is a description notifications that may be provided to
applications that maintain long-lived connections:
[0058] NetBT/RDR client: on some operating systems, a NetBT/RDR
client may attempt to connect using the NetBT protocol (over port
138/139) as well as using the SMB protocol (using port 445).
Connectivity using NetBT may be tried over all the available
interfaces and then the interface that connects first may be
picked. This is independent of the TCP/IP stack. Due to this
behavior, NetBT may actually end up picking a slower interface and
not taking the benefit of improvement in TCP/IP to select the best
interface.
[0059] To address this issue, in some embodiments the computer may
attempt to connect over port 445 first (which may end up using
TCP/IP routing preference), failing which it may fall back to
138/139 ports which tries connectivity over all interfaces. To do
this, SMB may prioritize TCP/IP connections by delaying issuance of
NetBT connections by 0.5 s.
[0060] NetBT may maintain long-lived connections which are shared
across multiple data transfers from/to the same host. This may
causes stickiness as to which interface gets picked even after a
computer (e.g., a laptop) is docked especially due to strong host
behavior since the routing is based on the source address as well,
which gets specified when the connection was initially established.
In addition, an operating system may implement a feature known as
fast reconnect which re-establishes a TCP connection after media
disconnect/connect (e.g., during sleep/wakeup) without tearing it
down. This means that a machine may be put in sleep mode and
connected via another interface before wake up may continue to pick
the previous interface independent of routing metric.
[0061] To address the long-lived connections issue, in some
embodiments a solution at the TCP layer may involve TCP migration
which is out of scope of this feature. A solution at the
application layer is possible however. SMB may perform the
following: [0062] 1. Register for media connect/disconnect
notifications [0063] 2. On a media connect notification; query the
stack if a better interface is available. [0064] 3. If a better
interface is available, tear down all connections that can be torn
down (i.e., no active file open) and that have been inactive for a
period of time
[0065] Having now described some embodiments of the invention, it
should be apparent to those skilled in the art that the foregoing
is merely illustrative and not limiting, having been presented by
way of example only. Numerous modifications and other embodiments
are within the scope of one of ordinary skill in the art and are
contemplated as falling within the scope of the invention. The
foregoing description and drawings are by way of example only. In
particular, although many of the examples presented herein involve
specific combinations of method acts or system elements, it should
be understood that those acts and those elements may be combined in
other ways to accomplish the same objectives. Acts, elements and
features discussed only in connection with one embodiment are not
intended to be excluded from a similar role in other
embodiments.
[0066] Use of ordinal terms such as "first", "second", "third",
etc., in the claims to modify a claim element does not by itself
connote any priority, precedence, or order of one claim element
over another or the temporal order in which acts of a method are
performed, but are used merely as labels to distinguish one claim
element having a certain name from another element having a same
name (but for use of the ordinal term) to distinguish the claim
elements. The use of "including," "comprising," or "having,"
"containing," "involving," and variations thereof herein, is meant
to encompass the items listed thereafter and equivalents thereof as
well as additional items.
* * * * *